Only God gets deference because magnifica humanitas is his creation and his alone. There have been many who have confused themselves into replacing God with their own image, only to have been forgotten, reviled, or otherwise replaced by the ticking of time, another invisible power that they did not bother to consult. Magnifica humanitas is fundamentally the human being in the individual and in the collective senses. The dignity conferred on the fingerprint is also the dignity of the masses. It is not to be confused with or displaced by humankind’s creations. Pope Leo XIV’s encyclical Magnifica Humanitas addresses this confusion and asks a few key questions (¶6): “Where are we going? Toward what goal do we wish to orient ourselves? What direction should we choose as a people and as a human community?”

To answer where we are going, let us start by looking at where we have been. Where we have not been is either utopia or apocalypse. Instead, we have historical patterns with recurring variables. Conversations about AI typically focus on attribution, accountability, equity, safety, or capability. This analysis starts with transmission, and, consistent with the encyclical and Christianity more generally, centers the solution on alignment to God.

Like the Bible itself, let us start with the most powerful transmission mechanism: the word. God speaks creation into existence, and in the Christian context, the Word is made flesh. With AI, math meets language. One could think of it as a human-built language deployed on a God-given language to produce a statistical distribution of language. In the case of the Tower of Babel, humans spoke one language deployed at one aim, self-glorification (Genesis 11:1–9). In the case of AI, there is not one target, but one average that might or might not fit all people or situations well. There are three primary actors with AI: the financier, the developer, and the user, and they interact to produce an output.

In terms of transmission, there are three primary mechanisms: administration, commerce, and the household. Administration can be thought of as the state. For example, the Roman Empire imposed Latin on its subjects, and it held until the empire did not, fracturing into the Romance languages. Setting aside the political factors that might also lead to one language’s dominance, commerce could be thought of as a lingua franca that is adopted to more easily facilitate economic transactions, with English being the most obvious example. Lastly, the household freely adopts and retains a language because of what it carries, the things it needs, even when the administration or commerce layer is absent. For example, the recovery of Hebrew as a spoken language to solidify a national identity was possible because the home and the synagogue preserved it for approximately two millennia without a state or a market.[1] In effect, the more freely a language is adopted, the more durable it is.

Working from this positive correlation between each human being’s exercise of free will as it relates to language and its viability, let us consider what an AI model is and is not. It does not on its own impose itself on its user as a state might, and it also has no free will as a person does. It makes no judgments and has no preferences of its own regarding language. To the degree that it facilitates commerce, it does so within the context of its time and place. AI is also not a human being; it has no anthropology and no soul. It is an artifact whose entire existence depends on humans, yet ontologically it is not, and will never be, in the same category. AI provides its user with competent answers to the question asked based on central tendency with movable conditioning determined by the user. A human being cannot be reconditioned during an exchange, but AI can be and is, by its developers in training, by the density of its corpus, and, importantly, by its users in every exchange.

Where are we going with AI? Let us consider Babel and separately the Golden Calf. Babel is one language directed by its people at self-aggrandizement, instead of God’s glory. Whatever harmony they held, their effort ends with God’s punishment in cacophony. The builders tried to erect a tower to carry their names to the heavens. In Babel, they were humbled by God. In our case, they will be humbled by God’s other creation, time. The Golden Calf is another form of misdirected worship (Exodus 32). After their true God delivered them out of bondage, the Israelites demanded that Aaron make them gods out of their gold, a golden calf, crediting it with unearned divine power. They were then punished by Moses and by God for violating the law they had received just weeks earlier, specifically worshipping false gods. Whether AI is used for the Tower of Babel or the Golden Calf, in either case, glory would not be directed to God.

Where should we be going with AI? Toward what goal do we wish to orient ourselves? What direction should we choose as a people and as a human community? Various normative responses are possible, but the Christian response would be to glorify God. In the Catholic context, this direction would be congruent with Catholic Social Teaching. Its practice would overlap with humanism in honoring the inherent dignity and value of each person and in keeping him or her unsubordinated to money, prestige, or power. In the Saint John’s Bible, Jesus Christ is represented in layers of gold so that the viewer can see his or her own face reflected back – the Imago Dei.[2] The gold is a symbol of the divine, and its value is not as a precious metal but as a mirror of God’s creation. AI can also mirror its users. The outstanding question is whether the image it produces helps or hurts them.

There are several factors that determine whether AI’s conditional outputs are beneficial or damaging to its users and to society. The most critical and the most overlooked covariate within the user’s control is the person’s own formation. To be clear, although the prompts do matter, teaching someone to provide better prompts cannot substitute for the decades of formation that teach a person how to think. Every model is designed to leverage central tendency, but the outliers are where the model’s sophistication lies and where the user can reveal its presence or absence. At the tail of the distribution, the model’s output quality correlates with the user’s input quality, which is determined by the person’s formation. The quality of the conversation or the product depends on the person’s self-regulation, clarity of thought, precision of language, domain knowledge, and comfort with critical thinking and feedback.

Other covariates outside of the user’s direct control include: the model’s “house style,” such as tone, which is influenced by raters and tends toward Western, educated, and secular; deployer instructions; guardrail activation; model bias; model size, with larger models sustaining tailored responses better and smaller ones reverting to generic scripts faster; corpus density, the thicker, the better; intra-sentence positioning and the chat’s length, with longer conversations more likely to drift; handling of random or confusing prompts; impetus to search or not; and token management. All these decisions are made by the developer or the deployer and can lead to common errors, such as hallucinations, fabrications, drift, sycophancy, confabulated reasoning, stale information, overconfidence even where the knowledge base is thin, pattern-matching overriding user and context logic (shortcut learning), limited cross-domain synthesis, and others.

Aside from the safety issues and the errors, AI’s greatest weakness is similar to the algorithmic flattening that has reduced distinctiveness since the rise of the social media feed. When AI generates or edits a text, it smooths, explains, excuses, cushions, moralizes, and pads, among other friction-reducing moves, which collectively have the effect of anonymizing a distinctive voice and patronizing the reader. AI can also inadvertently modify the meaning because it edits a part, such as a sentence, without tracking the style, structure, or substance of the whole. The intention is to make the text more polished, but good writing is produced by both what is visible and what is invisible, the countless judgments, including exclusions, that an author or editor makes. Being able to see the masterpiece in the rough stone and to bring style, structure, and substance into internal coherence by subtraction, what Michelangelo called per forza di levare, is as important as putting words on the page by addition, per via di porre.[3] AI is trained on what it can see. What it cannot see, the reasoning behind the subtraction, does not enter its training. The central tendency of AI models means that the voice of distinctive writers, the tails of the distribution, will be edited, by default, toward the mean.

AI serves magnifica humanitas when it helps bring the user’s fingerprint into relief in a way that benefits the child of God on the other side of the screen and the broader society we create. This goal can be clear, but the process to achieve it, like most aspects of building a society, is challenging. Achieving the goal cannot be outsourced to the AI companies. Like the dance between creator and created, God and human, there is also a dance between the model and the user, and each has to participate in the goal’s furtherance. A misaligned AI company can do great damage to the user and the society, and a poorly formed user will get a worse product than a well-formed one would. Although it inherits the same pretraining limits as closed-weight models, open-weight AI is one option to customize models to serve specific users or needs, support diversification, and reduce the dominance of Big Tech, the incumbent companies with structural advantages that both develop their own models and finance the frontier labs, such as Anthropic and OpenAI. Society needs to foster the formation of the individual and to create economic policies to advance smaller AI companies.

Focusing on the financier and developer side of AI, we can use Saint Benedict’s Rule as a template for responsible and effective governance. As all the saints were, he was oriented to God: “Prefer nothing to the love of Christ.” In secular terms, we would say to the common good. The Rule served as both a structure for living and a statement of values against which to measure the living. AI companies need to develop a written mandate that binds every stakeholder – company, labor, and customer – and serves the common good. Similar to the Rule, which was read often, the mandate needs to be revisited regularly to assess whether the company is adhering to it. All stakeholders’ voice should be heard and, importantly, truly listened to, and they should be required to be consulted before any major company decisions are made. The leaders who make the final decisions need to be evaluated for their ability to deliver both performance for the company and adherence to the mandate. Stock performance should not be the primary metric. Instead, long-term performance and adherence to the mandate should be the primary metrics. As with the Benedictine communities, which were quite financially successful, stability should be balanced with productive change. The governance structure should protect the company from capture or takeover. Lastly, there need to be caps on executive compensation and reasonable ratios relative to rank-and-file workers, and any surpluses should be reinvested for the long-term interests of the company and in fairly compensating labor, and serving its customers and the broader society.

The household level, the one on which society is the most dependent for transmission, is where the failures have been the most damaging. Families have stopped forming their children as they once did, and, intentionally or unintentionally, the technologies their children use have become part-time parents and, in some cases, full-time regulators. Instead of the child developing a body-mind relationship and self-regulation, a foreign object, a machine, with or without AI, has become an intermediating party. The result is predictable. The child is poorly formed. In 2024, children under two averaged nearly one hour of screen time a day and two- to four-year-olds about two hours (Common Sense Media, 2025), which is correlated with disrupted sleep and development. Sixty-eight percent of parents say their phone distracts them at least sometimes when they are with their children (Auxier et al., 2020). To get a sense of the magnitude of the problem, in 2025, the average for Americans was about 2.6 hours of TV (BLS (a), 2026) and about 2.3 hours of social media a day (Statista, 2025), which is, allowing for overlap between devices, roughly 30 hours a week. Over a lifetime, that is about 19 waking years, or more than seven times the hours of all of K-12.[4] Conversely, reading gets about 16 minutes a day (BLS (a), 2026) and religious practice about 8 minutes (BLS (b), 2026). Formation is slow and has to be intentional. If this is intentional, it is misguided. If it is unintentional, the costs of the ignorance are high. When a generation’s greatest inheritance is an addiction that has warped their formation, the adults and the society owe them an explanation and a remedy.  

The problem with our current society is that companies, institutions – including the church – and households are all failing simultaneously. Part of this is related to the secularization of the culture, driven by its own forces and exacerbated by the church’s own failures, including but not limited to the child sex abuse scandals on the most egregious end and clericalism, such as focusing on personal ambitions and ego instead of providing pastoral care, on the more widespread one. Academia has focused more on itself than on the society it claims to serve. The scholarship it produces is more for vainglory and vanity, or even knowledge for its own sake, than for the common good, and the country has noticed. Many of the heads of companies, and with them the companies they lead, have become vehicles for self-aggrandizement. For example, Elon Musk would do well to remember that “Elon Musk” the name would not exist without the United States, and Elon Musk the human being would not exist without God. The public would do well to remember its own power. The common thread is Babel or the Calf.

Only God gets deference. The rest of us get accountability. Like Nehemiah, a Jew called after the Babylonian Exile to rebuild the holy city of Jerusalem, all members of our society are called to be repairers of the breach (Nehemiah 2:5; Isaiah 58:12). The damaged wall we face is our responsibility to restore. While we no longer speak a common language, Pentecost reminds us that if we align ourselves to God, by the grace of the Holy Spirit, we will understand each other (Acts 2:1–11). Instead of raising a Tower of Babel or elevating a Golden Calf, let us uplift each other in a democratic spirit of solidarity. The household is the most enduring transmission mechanism. A nation’s people hold its purse. The power of both can mitigate the excesses and abuses of the state or the market. For a bottom-up approach to be successful, however, it has to keep its eyes fixed on something higher than itself and past its own time, into the generations that will come after it and hold it accountable if only in the history books. We all have a fingerprint and a name known to God. What gets written about us as a people should honor the spark of the divine with which he endowed us.

by Ann Job, published on October 20, 2026

---

Notes

[1] Britannica, “Hebrew language,” britannica.com/topic/Hebrew-language, updated October 1, 2026.

[2] The Saint John’s Bible, commissioned by Saint John’s Abbey and University, Collegeville, Minnesota, Donald Jackson, artistic director, completed 2011. The reading of the gold as a mirror is from Tim Ternes, curator of The Saint John’s Bible, Hill Museum & Manuscript Library, presentation at Hill-Murray High School, October 5, 2026.

[3] Tom Gurney, “Michelangelo Quotes,” TheHistoryOfArt.org, thehistoryofart.org/michelangelo/quotes, updated November 6, 2025. “Io intendo scultura quella che si fa per forza di levare; quella che si fa per via di porre è simile alla pittura” (letter to Benedetto Varchi, 1547).

[4] At roughly 30 hours a week from age 10 to 80, screen time totals about 109,200 hours (30 × 52 × 70). Assuming 16 waking hours a day (5,840 a year), that equals about 18.7 waking years. K-12 schooling, at 13 years of about 180 days and 6.5 hours a day, totals about 15,210 hours, so lifetime screen time is roughly 7.2 times all of K-12. See the National Center for Education Statistics, which reports the typical US school year at about 180 days and the school day at about 6.5 hours.

Sources

Auxier, Brooke, Monica Anderson, Andrew Perrin, and Erica Turner. Parenting Children in the Age of Screens. Pew Research Center, July 28, 2020. pewresearch.org/internet/2020/07/28/parenting-children-in-the-age-of-screens.

Common Sense Media. The 2025 Common Sense Census: Media Use by Kids Zero to Eight. February 26, 2025. commonsensemedia.org/research/the-2025-common-sense-census-media-use-by-kids-zero-to-eight.

National Center for Education Statistics. “Average number of hours in the school day and average number of days in the school year for public schools, by state: 2007–08.” Schools and Staffing Survey (SASS). nces.ed.gov/surveys/sass/tables/sass0708_035_s1s.asp.

Statista. “Daily time spent on social networking by internet users worldwide from 2012 to 2025 (in minutes).” Updated June 19, 2025. statista.com/statistics/433871/daily-social-media-usage-worldwide.

U.S. Bureau of Labor Statistics (a). “Americans age 15 and over read for an average of 16 minutes per day in 2025.” TED: The Economics Daily, July 22, 2026. bls.gov/opub/ted/2026/americans-age-15-and-over-read-for-an-average-of-16-minutes-per-day-in-2025.htm.

U.S. Bureau of Labor Statistics (b). “Table 1. Time spent in primary activities and percent of the civilian population engaging in each activity, averages per day by sex, 2025 annual averages.” American Time Use Survey, Economic News Release, June 25, 2026. bls.gov/news.release/atus.t01.htm.

Anthropic’s Claude served as a research assistant and copy editor. The ideas, the argument and the writing are distinctively my own, including the embedded theory of language transmission: state, commerce and household, and the positive correlation between free adoption and durability.